We introduce two institutions that provide multiple public good units, assuming that a market-maker has the ability to establish groups of contributors. We set up a public good experiment where either all N individuals form one group to provide two units, or divide the N participants into two groups, and each group provides one unit separately, with all individuals benefiting for any unit(s) provided. Our results show that, under certain circumstances, the latter approach provides more of both units, and it encourages more contribution on average. We also explore the performance of two rebate rules under the two grouping approaches.